Acute myeloid leukemia (AML) is a malignancy that is characterized by molecular and clinical heterogeneity. Changes to the classification of AML over the last decades are representative of our evolving understanding of cancer biology and treatment approaches in general. Initially classified using morphologic criteria, advances in cytogenetic and now genomic sequencing have resulted in a better delineation of various AML subtypes according to pathologic features.
It has long been known that recurrent cytogenetic abnormalities have an important role in prognostication and contribute to treatment decisions. For instance, patients with AML characterized by balanced translocations in core binding factors tend to have disease that is curable with chemotherapy alone, while patients with complex cytogenetic abnormalities have a poor prognosis and generally are recommended for stem cell transplantation. Beyond cytogenetic abnormalities, in recent years the genomic mutational landscape of AML has been characterized, identifying a number of recurrent somatic mutations in AML, such that nearly all cases of leukemia may be defined according to their mutation profile . Recurrent somatic mutations in AML have been instrumental in better understanding the pathogenesis of the disease (e.g., how sequential somatic mutations may cooperate to drive leukemogenesis) .
In addition, a greater understanding of recurrent mutations in AML has led to the development of a number of therapeutics targeting these specific abnormalities. Most notably, the identification of internal tandem duplications and tyrosine kinase domain mutations in the FMS-like tyrosine kinase 3 receptor (FLT3-ITD and FLT3-TKD) has generated clinical investigation into FLT3 inhibitors, including the approval in 2017 of the tyrosine kinase inhibitor midostaurin when added to standard AML induction chemotherapy . More recently, a selective inhibitor of mutated IDH2, enasidenib, was approved for the treatment of relapsed or refractory IDH2- mutated AML , and there is encouraging data for a similar agent, ivosidenib, in IDH1-mutated disease.
These gains are reliant upon not just the development of new therapies but also upon the development and dissemination of reliable testing methods and treatment options to clinicians treating AML patients. Successes with targeted therapies in clinical trials may not translate into population health gains if patients are not tested for these mutations. As such, understanding the type of testing that an AML patient has at diagnosis, and how that influences patient outcomes, is of interest.
One way to identify testing patterns is to look at how AML cases are reported to cancer registries. Mirroring the gains made in AML pathobiology over the past decades, classification systems have similarly undergone updates that are representative of our understanding of the diagnostic and prognostic heterogeneity of the disease. In contrast to the FAB classification, the WHO classification and revisions have increasingly incorporated new prognostically relevant subtypes of disease according to the best available data at that time; as such, the way AML is reported, whether by FAB or by various WHO categories, may be an indicator of the local oncologic practices and may translate into patient outcomes.
This is the question that Song et al.  sought to understand in this issue of Acta Haematologica: whether how AML is classified (i.e., FAB classifications, updated WHO criteria, or histologically unclassified) impacts patient outcomes. They chose to interrogate the United States National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) database, which is a population-based cancer registry that covers approximately 29% of the US population and is broadly representative of the US population as a whole. SEER captures incident cancer cases as well as survival with high fidelity; the data that SEER distributes is accurate to the degree that it is reported to each cancer registry. As such, incident AML cases that do not have molecular data reported, or pathology reports that do not delineate a specific WHO subgroup, may be classified as AML not otherwise specified in SEER.
Song et al.  evaluated how survival varied depending on whether AML patient cases were reported according to FAB groups, WHO criteria, or unspecified subtypes. They identified over 32,000 patients diagnosed with AML between 2001 and 2013, and only approximately half had AML reported to SEER according to a WHO classification. This did improve over the study period; nonetheless, the percentage of patients with a histological AML NOS classification remained high, even in 2013. Moreover, this study found that patients who were reported as AML NOS in SEER, as a histological classification rather than the WHO subgroup, had a significantly worse survival (median survival 33.7 months, compared to 41.3 months for patients with WHO classification). There was also variation in outcomes according to race; black patients had a worse survival compared to white AML patients, and black patients also were more likely to have AML NOS due to histological classification (9%) than being in the WHO AML NOS subgroup (7%). This suggests that black patients are not having optimal diagnostics performed at AML diagnosis, and this may contribute to the worse survival seen in this group.
These findings suggest a number of things. First, assuming unclassified AML is not enriched for higher risk subtypes, it suggests that classification of AML using the most current standards is correlated with improved outcomes. We do not know whether this reflects a better overall care of AML patients or if this is indicative of the practice patterns of cancer centers according to volume or AML specialization, but it appears to be a marker. Second, the uptake of new classification systems is slower than would be desired, with only half of patients classified according to WHO criteria. As novel AML therapies are developed that are contingent upon the presence of a specific mutation, it will be critical that patients have such testing in the future, and these data suggest that we need to disseminate such testing more effectively.
Approximately half of AML cases do not get reported according to the current WHO classification, and Song et al.  showed that this is associated with a worsened overall survival in AML patients. This paper represents a call for improved diagnostics in cancer care. As a field we must ensure that we have cytogenetic and, increasingly, complete molecular testing to fully characterize AML diagnoses, so that we may translate new treatment modalities and survival benefits into the larger population.